AI Model Learns Urban Access Costs from Mobility Data

This research addresses a common challenge in urban planning: understanding the hidden costs that influence how people access services within cities. While planners often observe where people go, the underlying factors like distance, price, and institutional access that drive these choices remain unclear. The study proposes an inverse optimal transport framework to recover these latent choice costs from observed origin-destination flows. Using school-to-school enrollment data in the Philippines as a case study, the researchers applied two models: an interpretable distance-banded model with a subsidy term, and a neural cost model. Both models utilize a differentiable Sinkhorn forward pass to train. The framework successfully estimated a "subsidy-equivalent distance," which quantifies the perceived travel cost offset by an education subsidy. This demonstrates how administrative mobility data can be transformed into actionable planning metrics, aiding in the design of accessibility-aware subsidies, optimal facility siting, and efficient urban service allocation.